The high count rates given by the ME required OBC (on board computer) programs to compress the data prior to their being telemetered. Depending on the objective of the observation the OBC programs traded time resolution against spectral information. Depending on the telemetry load, and the OBC programs running for the other two experiments, two or three ME programs could be run simultaneously. The spectral orientated programs gave spectra plus intensity profiles. The timing programs gave purely intensity profile data with in some cases selectable channels. The highest time resolution possible for a single selectable energy band was 0.2 ms.
The products available within this database has been created using the data sampled by the spectral orientated OBC programs.
Spectra: For each observing interval there is a background-subtracted on-source spectrum integrated over the entire observing interval. The background subtraction method to derive the on source spectra depends on the mechanical configuration of the ME array detector used during an observation. The quality of these spectra is primarily determined by how good the background subtraction was. This is indicated by the QFLAG parameter which has a range from 1 to 5. Spectra with a QFLAG less than 3 are usually not of sufficient quality and should not be used. pTo each entry in the database is associated a spectral file in FITS format and a plot of the FITS data in GIF format.
Lightcurves: Typically during an EXOSAT observation, one or a combination of the following mechanical ME configurations were used:
* only one half of the ME array was on source and the other half monitoring the background throughout an observation * both half on source * the two halfs alternatively were monitoring the source and the background swapping after few hours.For each mechanical configuration interval the following type of lightcurves in FITS format are available:
* On-source background-subtracted 30 s light curves in the energy range 1-8 keV. The filename of these lightcurves starts with the suffix 'D'. * Background light curves in the energy range 1-8 keV with a time resolution of 30 s (non available for co-aligned configuration). The filename of these lightcurves starts with the suffix 'R'. * Multiband (1-10 s) time resolution on-source background-subtracted light curves in the energy bands 1-3.8, 3.8-8 and 1-8 keV. The filename of these lightcurves starts with the suffix 'A', 'B' and 'C' respectively. The time resolution of these files is the maximum available from the pulse height analyzed (spectral) data and depends on that specified by the principle investigator.Lightcurves were accumulated everytime a new mechanical configuration occurred during an observation interval and therefore one or more FITS lightcurves, for each type, can be associated to a database entry.
The lightcurves are also provided as GIF files. To each entry in the database are associated two GIFs. One shows the full energy band (0.8-9 keV 'D' files), the other shows the lightcurves for the 0.8-3.6 keV band and 3.6-9 keV band (the 'A' and 'B' files). The lightcurves plotted in the GIF file are obtained using all the files associated to that database entry.
The ME auto-processed only the argon chamber data. It proceeded as follows:
* Selection of the method of background subtraction, which depends on the set up of the ME configuartion during the observation. * Creation of spectra and lightcurve files. * Spectral and timing analysis.The selection of the most appropriate method of background subtraction for a particular observation is the crucial step in determining the quality of the data products and results. The auto chooses in a hierarchical manner from the following list of available background subtraction techniques, using:
* Method 1: the generation of difference spectra (SWAP3A), only possible for observation in which the two halfs of the ME array alternatively were monitoring the source and the background swapping after few hours. * Method 2: the standard difference spectra (SWAP1,SWAP2,SWAP2A or SWAP3B), possible only as in Method 1 a swap is performed during an observation. * Method 3: slew spectra (SLEW1 or SLEW2), typically used when the same half or both half are on source thoughout an observation. * Method 4: standard backgrounds (NOSLEW), used only if slew data are not available or contaminated by a source.Methods 1 and 2 generally produce the best results because background data is obtained from the same detectors at a different time when they were offset to view a source-free region of sky. The major uncertainty in these cases is due to a difference in the background spectrum and counting rate between the aligned and offset positions. This difference is removed using "difference spectra".
In method 1 the array swaps were done in a sequence where the difference spectra can be uniquely determined from +ve and -ve offset background spectra (see EXOSAT Express No. 16, p. 21). This usually gives the best background subtraction possible. If the set of array swaps was incomplete then method 2 uses standard difference spectra generated from background observations. If no suitable array swap data are available then method 3 uses background spectra obtained during the slew on to or off from (or both) the source. If no suitable slew data are available, then method 4 uses standard background spectra. Method 4 usually produces satisfactory results only for bright sources.
If the background subtraction was unsatisfactory after the first pass a number of "fixes" were tried:
* Time windows were used to exclude times when the background was flaring. * If satisfactory background subtraction cannot be obtained using methods 1 and 2 then the same macros are rerun using only the corner detectors. These have very small (or zero) difference spectra. If this procedure has been followed it can be, when plotting a spectrum, that dets = 10011001. * If one detector appears to be significantly noisier than the others, then it is excluded from the analysis. * If method 4 has been used for background subtraction (NOSLEW) then an attempt is made to find a more suitable "standard" background. The default backgrounds are averaged over the whole mission and sometimes better backgrounds can be found from nearby observations.In addition, a number of other problems were dealt with:
* Occasionally there was a temporary loss of spacecraft pointing during an observation. These intervals are excluded from the auto. * Sometimes the automatic software would assume that the pointing position of the previous observation applied to the current one. This is corrected by cross-correlating the results database with the EXOSAT log. * Any bursts (type I only) are removed from the spectra (but not the light curves). A comment is then added to the spectrum to indicate this.After selecting the background subtraction method and fixing the other problems a spectrum was created integrated over the entire observation from all detectors. In addition, light curve files are created at the highest time resolution available in the spectral telemetry stream; this typically ranges between 1 and 10 s, although on occasion higher time resolution was available. Three files are created with light curves covering the energy bands of 1--3, 3--8 and 1--8 keV for the source and 1--8 keV for the background data.
If the source count rate is > 0.25 ct/s/half, then automatic spectral fitting and timing analysis was carried out, with the results written to the database.
Qflag Meaning 0 unusable 1 very poor 2 poor 3 acceptable 4 very good 5 excellentGenerally, spectra with quality flags of 3 or more are suitable for spectral analysis while those with quality flags of 0, 1 or 2 have major problems and should only be used with extreme caution. In many cases it may be possible to improve the quality of the background subtraction by returning to the FOTs and carrying out a more complex analysis.
There are a number of points that the user should carefully check before using ME auto products and results.
The ME quality flag that is assigned to each entry is a subjective measure of how well the background subtraction has been done. It does not take into account the possibility of there being more than one source within the 45 arc minute ME field-of-view --- such entries can still have good quality flags and it is up to the user to determine from where the observed flux originates.
Another potential problem is the presence of weak contaminating X-ray sources in the offset detectors. Their effect will depend on the direction of any array swaps and the background subtraction method used, but will generally cause the background to be over-subtracted. Plotting the background counting rate `pp/bg` is a good way of checking for this.
When plotting the background counting rates from the offset detectors there will always be a step up or down in counting rate at the time of an array swap. This occurs because the background is slightly different in the two detector halves.
Lastly, after a slew on to a source there was always an interval of usually a few minutes, but sometimes up to an hour, before the pointing control was switched to the star trackers. When this occurs the pointing sometimes changed by several arc minutes and this can cause a step up in counting rate. These should have been removed in the quality checking, but it is possible that some have been missed.
The spectral parameters obtained from the automatic spectral fitting should be used with great caution. Automated spectral fitting programs are far from infallible and in many cases may have gone astray. The following problems may arise:
* there may be more than one model that gives an acceptable fit * the automatic spectral fitting program sometimes was trapped in a false minimum, causing apparently no model, or the wrong model, to give the best fit. This occurs most often for models 6 -- 8. * entries with `Qflag ME` <3 indicate problems with background subtraction, which may affect the spectral fit.Always extract the spectrum and fit again using XSPEC.
* All observations after day 77 of 1986. After this time the EXOSAT pointing was not stable and it is necessary to apply a time-dependent collimator correction to light curves and spectra. * Some of the very early observations where there is no useful background spectrum.
In the case of the light curves, the count rates have been corrected for collimator transmission using the source position given by parameters `RA` and `Dec` and spacecraft pointing position in `point RA` and `point Dec`.
Other count rates and associated errors are
* `Rate 1` - The count rate in band 1, 1-3 keV * `Rate 2` - The count rate in band 2, 3-6 keV * `Rate 3` - The count rate in band 3, 6-10 keV * `Rate 4` - The count rate in band 4, 10-15 keV * `Rate total` - The count rate in the entire available band
HEASARC User Hotline.